Skip to content

Releases: hi-ogawa/demucs-onnx

Demucs ONNX models (models-2026-07-11)

Choose a tag to compare

@hi-ogawa hi-ogawa released this 11 Jul 11:52
0625df8

Size-optimized Demucs ONNX models for native and browser inference. These are derived artifacts based on the upstream Demucs models and the ONNX export work in adefossez/demucs#10.

Download one of these sets:

  • Standard four-stem and two-stem modes: dft.bin and htdemucs.onnx.
  • Fine-tuned target/minus mode: dft.bin and the corresponding htdemucs_ft_<source>.onnx specialist.
  • Full fine-tuned support: dft.bin and all four fine-tuned specialists.

Every ONNX model references the shared dft.bin external-data file. Keep both files in the same directory for native ONNX Runtime, or supply dft.bin through ONNX Runtime Web's external-data API.